Eighty-one immunoscintigraphy (IS) studies in 80 patients with gastrointestinal malignancies were re-evaluated retrospectively to analyze accuracy and clinical impact of the IS findings. Selection of patients was based upon complete diagnostic records and a clinical follow-up of at least 6 months. The 3 antibody preparations used (99m-Tc labelled complete anti-CEA antibodies (BW 431/26), 111-In labelled F(ab')2-fragments against CEA (BW 431/31), and a mixture of 131-I labelled F(ab')2-fragments against CEA and CA 19-9 (IMACIS-1] yielded equal diagnostic sensitivities (65%, range 60-78%) except for liver metastases. These were preferably detected by the 131-I labelled compound. Best results were gained in local recurrencies of gastrointestinal cancer (12/15 true positive = 80%). Among these were 3 out of the 4 lesions which had been exclusively recognized by IS. Thus IS should be applied in patients with suggested local recurrencies and inconclusive outcome of routine diagnostic workup. A widespread and frequent use of IS (e.g. screening of metastases) cannot be advocated since diagnostic results were not convincing and the immunologic risks of IS (development of human anti-mouse antibodies) are still under discussion.